Mechanisms of neurovascular regulation in health and disease: Insights from experimental animal models

Neuronal signaling mechanisms

Neuronal activity initiates NVC through a variety of parallel signals that engage different cellular targets within the neurovascular unit. These include direct modulation of vascular tone via diffusible messengers and specialized junctions, as well as indirect pathways that recruit astrocytes, endothelial cells, pericytes, and microglia, each of which is addressed in detail in subsequent sections. Briefly, although still debated, glutamate released during synaptic activity may act on astrocytic metabotropic glutamate receptors, triggering intracellular Ca²⁺ elevations and the subsequent production of vasoactive mediators. ³³ Glutamate may also directly influence vascular tone by acting on NMDA receptors expressed on both VSMCs ³⁴ and ECs.^35,36 ATP released by activated neurons can activate purinergic receptors on astrocytes, leading in turn to prostaglandin production and vasodilation via pericytes. ³⁷ Adenosine, produced by the breakdown of extracellular ATP, can act directly on VSMCs via adenosine receptors ³⁸ or activate K_ATP channels on ECs and pericytes. ³⁹ The release of cyclooxygenase-2 (COX-2)-derived prostaglandin E₂ (PGE₂) from pyramidal neurons has also been proposed as a mechanism by which these cells contribute to NVC.^40,41 Finally, ionic changes evoked by neuronal activity, particularly increases in extracellular K⁺, are critical for initiating endothelium-dependent, retrogradely propagated vasodilation, a key component of the CBF response. ⁴²

However, the most characterized mechanism by which neurons appear to modulate CBF is through the release of nitric oxide (NO). Activation of postsynaptic NMDA and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, leading to elevations in cytoplasmic Ca²⁺ driven by Ca²⁺ influx through ion channels and Ca²⁺ release from internal stores. The resulting rise in cytosolic Ca²⁺ activates Ca²⁺-dependent enzymes, most prominently neuronal nitric oxide synthase (nNOS) and cyclooxygenase-2 (COX-2), which produce potent vasodilators, namely NO and prostanoids, respectively. ⁴ NO is uniquely suited to serve as a rapid initiator of functional hyperemia. As a membrane-permeant gas, NO diffuses across cellular membranes to reach VSMCs, where it activates soluble guanylyl cyclase, elevates intracellular cGMP levels, and promotes relaxation through reduction of intracellular Ca²⁺ and desensitization of the contractile apparatus ⁴³ (Figure 2). A recent systematic analysis of pharmacological and genetic manipulations targeting proposed NVC signaling pathways identified nNOS-derived NO as the most critical mediator for the development of the neurovascular response. ³² Optogenetic stimulation of interneurons evokes larger vasodilation than stimulation of pyramidal neurons,^44–48 suggesting that a relatively small subset of neurons plays a disproportionately prominent role in regulating vascular tone. Among interneurons, the stimulation of those expressing nNOS produces robust increases in blood flow and vasodilation,^48–51 further underscoring that neuron-derived NO is a key regulator of NVC. However, the contribution of NO to NVC is region-specific. In the cerebellum, functional hyperemia is profoundly attenuated in nNOS knockout mice,^52–54 indicating a dominant role for neuronal NO. In contrast, cortical and hippocampal responses are only partially reduced,^54–57 suggesting that NO function as modulatory or permissive factor rather than the sole initiator factor in these regions. ⁵⁷ Thus, while NO represents a critical initiator of NVC, its relative contribution varies across brain regions.

OPEN IN VIEWER

Figure 2.

Mechanisms of neurovascular coupling at the arteriolar level. Activation of NMDARs in postsynaptic neurons leads to an increase in intracellular Ca²⁺, which in turn activates neuronal NO synthase, promoting the synthesis and release of NO. NO then activates sGC in VSMCs, resulting in increased cGMP levels and VSMC relaxation. Glutamate released by activated neurons may also activate NMDARs expressed on the abluminal side of ECs, leading to increased Ca²⁺ levels, activation of endothelial NO synthase, and subsequent VSMC relaxation. Glutamate released at the NsMJs activates NMDA on VSMCs, causing Ca²⁺ influx, activation of BK_Ca channels, VSMC hyperpolarization, and relaxation. Increased Ca²⁺ levels in astrocytes induces the release of arachidonic acid, which is converted to PGE₂ and EETs, both of which act on VSMCs to promote relaxation. ATP released by activated neurons is converted to ADP by the ectonucleotidase CD39 in microglia; ADP is then further converted to adenosine, which acts on A₂R on VSMCs, inducing relaxation. Endothelial retrograde hyperpolarization is propagated through gap junctions connecting ECs to upstream segments of the cerebrovascular tree and may also hyperpolarize VSMCs via gap junctions at myoendothelial junctions. Dashed lines indicate pathways for which evidence is limited.

Beyond its direct vasodilatory action, NO also exerts important regulatory effects on arachidonic acid (AA) metabolism. NO suppresses the synthesis of 20-hydroxyeicosatetraenoic acid (20-HETE), a potent vasoconstrictor derived from AA that inhibits large-conductance Ca²⁺-activated potassium channels (BK_Ca) in VSMCs. ⁵⁸ During neuronal activation, NO may therefore appear necessary both to restrain this constrictor pathway and to offset inhibitory effects of 20-HETE on BK_Ca channels. ⁵⁸ By limiting 20-HETE synthesis, NO may also facilitate vasodilation mediated by other AA-derived factors, including epoxyeicosatrienoic acids (EETs), which activate BK_Ca channels, promote membrane hyperpolarization and enhance VSMCs relaxation. ⁵⁸ Collectively, these observations exemplify how, in certain brain regions, NO appears to function as a permissive mediator within the neurovascular signaling network, modulating the balance between vasoconstrictor and vasodilatory pathways rather than serving as the driver of functional hyperemia. Regional differences in the density of nNOS-expressing interneurons may contribute to variability in the extent to which NO assumes this permissive role. Consistent with a modulatory rather than exclusively primary function, recent human data demonstrate that selective nNOS inhibition alters the kinetics of the initial NVC response without significantly affecting peak amplitude. ⁵⁹ These findings suggest that neuronal NO is particularly important for shaping the rapid onset of functional hyperemia, whereas sustained dilation likely depends on additional pathways.

In addition to diffusible NO, specialized neural-arteriolar smooth muscle cell junctions (NsMJs) ³⁴ provide a rapid and spatially precise conduit for glutamatergic signaling to arteriolar smooth muscle. Glutamate released by activated neurons directly engages NMDA receptors located on the abluminal membrane of arteriolar smooth muscle cells (aSMCs), triggering vasodilation ³⁴ (Figure 2). NsMJ-mediated transmission may account for up to ~60% of the CBF increase evoked by sensory stimulation under physiological conditions. ³⁴ Mechanistically, NMDA receptor activation induces Ca²⁺ influx in aSMCs, which activates BK_Ca channels, increases K⁺ efflux, promotes membrane hyperpolarization and results in smooth muscle relaxation. ³⁴ Although compelling, these findings require further replication across brain regions and stimulation paradigms to define their generalizability and quantitative contribution relative to NO-dependent signaling.

Endothelial electrical conduction

A major conceptual advance in NVC has been the recognition that capillary endothelial cells (cECs) function as electrically coupled networks capable of retrograde signal propagation.^42,60–63 Chen et al. ⁶⁰ first demonstrated that, similar to peripheral ECs, brain ECs can initiate retrogradely propagated vasodilation, that is, essential for the dilation of upstream segments of the cerebrovascular tree in response to neural activity. ⁶⁰ Subsequent studies established that brain cECs act as sensors of local neuronal activity throughout the brain, supporting a mechanism in which the continuous, interconnected capillary network functions as a system of “wires” that transmit electrical signals to upstream parenchymal arterioles and ultimately to pial arteries. ⁴² Specifically, extracellular K⁺ released during neural activity activates inwardly rectifying potassium channels (K_IR2.1) on cECs. ⁴² This produces rapid hyperpolarization that spreads from cell to cell via endothelial gap junctions and is transmitted to adjacent VSMCs at myoendothelial junctions, leading to smooth muscle hyperpolarization and arteriolar dilation. ⁴² Because K_IR2.1 channels are themselves activated by hyperpolarization, this signal regenerates along the capillary network, enabling rapid retrograde conduction over substantial distance, ultimately reaching upstream arterioles and pial arteries ⁴² (Figures 2–4). Consistent with this hypothesis, endothelial deletion of K_IR2.1 blocked the propagation of the vasodilation and suppressed the increase in CBF produced by whisker stimulation. ⁴² K_IR2.1 channel function in cECs requires phosphatidylinositol 4,5-bisphosphate (PIP₂), a phospholipid localized in the endothelial plasma membrane critical for ion channel regulation. ⁶⁴ Consistently, degradation of PIP₂ impairs K_IR2.1 function and activates capillary TRPV4 channels, thereby dampening retrograde electrical signaling. ⁶⁵ A recent study has suggested an additional mechanism by which cECs may initiate retrograde hyperpolarization. ⁶²

OPEN IN VIEWER

Figure 3.

Mechanisms of neurovascular coupling in capillaries (first to fourth branch). ATP released during neuronal activity activates P₂X receptors on astrocytes, leading to increased intracellular Ca²⁺ levels. The Ca²⁺ increase induces the release of AA, which is subsequently converted to PGE₂ and EETs. Both mediators act on mural cells to promote relaxation. Specifically, PGE₂ activates EP₄ receptors, whereas EETs induce hyperpolarization of pericytes, resulting in pericyte relaxation. Increases in Ca²⁺ levels in astrocytes can also activate BK_Ca channels, leading to K⁺ efflux. The resulting rise in extracellular K⁺ activates K_IR channels in VSMCs, causing VSMC hyperpolarization, and ultimately vasodilation. Increases in extracellular K⁺ following neuronal activation activate K_IR2.1 channels in ECs, causing membrane hyperpolarization. Because K_IR2.1 channels are themselves activated by hyperpolarization, this electrical signal is regenerated and rapidly propagates retrogradely from cell to cell through the capillary network via gap junctions, ultimately reaching upstream arterioles and pial arteries. PIP₂ is required for K_IR2.1 activity in ECs. Dashed lines represent pathways for which evidence is limited. ATP released by activated neurons is also converted to ADP by the ectonucleotidase CD39 in microglia and then to adenosine contributing to VSMCs relaxation and vasodilation.

OPEN IN VIEWER

Figure 4.

Capillary-initiated electrical signals driving upstream arteriolar dilation. Neuronal activity increases extracellular K⁺, activating K_IR2.1 channels in capillary ECs, and inducing membrane hyperpolarization that rapidly propagates retrogradely through gap junctions connecting adjacent ECs within the capillary network. This regenerative electrical signal spreads to precapillary and parenchymal arterioles, where it is transmitted to VSMCs and pericytes through cell-cell junctions, inducing their hyperpolarization of these cells and subsequent vasodilation. In parallel, TRPA1 channel-mediated Ca²⁺ influx in capillary ECs triggers ATP release through Panx1 channels, which activates purinergic P₂X receptors on adjacent endothelial cells, generating Ca²⁺ signals that more slowly propagate retrogradely contributing to upstream arteriolar dilation. Although the molecular mechanisms remain to be fully elucidated, thin-strand pericytes can sense neuronal activity and convert it into K_ATP-dependent electrical signals that propagate retrogradely to upstream arterioles, also contributing to the initial cerebral blood flow increase associated with neuronal activity.

An additional endothelial initiation pathway involves transient receptor potential ankyrin 1 (TRPA1) channel. ⁶² TRPA1 activation in cECs generates Ca²⁺ signals that propagate more slowly through the capillary network and induces dilation of upstream arterioles via endothelial pannexin-1 and purinergic signaling ⁶² (Figure 4). TRPA1-mediated signals propagate at approximately half the speed of K_IR2.1-mediated conduction, indicating distinct mechanistic pathways. ⁶² Because reactive oxygen species (ROS) can activate TRPA1 channels, ⁶⁶ it has been proposed that extracellular ROS generated by astrocytes or active neurons near brain capillaries may serve as physiological triggers. ⁶²

Electrical conduction within arterioles depends on gap junction coupling among ECs, among VSMCs, and at myoendothelial junctions. ¹² Arterial endothelial gap junctions critically determine both the speed and spatial extent of vasodilation propagation evoked by neural activity. ⁶³ Gap junctions are formed by membrane proteins called connexins, which are highly expressed in arterial endothelial cells (aECs). ⁶³ In Cx37/Cx40 double-knockout mice, stimulus-evoked vasodilation decays sharply with distance from the activation site, resulting in a blunted and spatially constrained dilation response. ⁶³

Additional evidence implicates caveolin and endothelial NO synthase (eNOS) in NVC regulation. Genetic deletion of either protein attenuates arteriolar dilation following whisker stimulation, suggesting that endothelial NO signaling contributes to full expression of the hyperemic response.^61,67 Brain ECs also express neurotransmitter receptor subunits that can be directly targeted by neurotransmitters.^35,36 Consistent with this, endothelial downregulation of the NMDA receptor subunit GluN1, mainly expressed on the basolateral endothelial membrane, suppresses NVC^35,36 (Figure 2). Finally, brain ECs may not only contribute to the initiation of NVC but also participate in the return of blood flow to baseline through a mechanosensitive feedback system, in which activation of endothelial Piezo1 channels accelerates the recovery of CBF at the conclusion of neural stimulation. ⁶⁸ Collectively, these findings position brain ECs as central initiators and propagators within the neurovascular unit.

Astrocytes: Amplification and sustained CBF response

While neuronal NO and specialized neural-arteriolar junctions provide rapid initiation of neurovascular coupling (NVC) and capillary endothelial cells propagate electrical signals retrogradely to upstream arterioles, astrocytes appear to contribute primarily to amplification and maintenance of the CBF response. The role of astrocytes in NVC was first demonstrated by Zonta et al., ³³ who showed that neuronal activity activates astrocytic metabotropic glutamate receptors (mGluRs), triggering intracellular Ca²⁺ release and inducing arteriole dilation both ex vivo and in vivo. Direct stimulation of astrocyte Ca²⁺ signals or mGluRs was sufficient to drive vasodilation, largely dependent on cyclooxygenase (COX) products, most likely astrocyte-derived prostaglandin E₂. ³³ Subsequent studies in brain slices^69,70 and in vivo ⁷¹ confirmed that astrocyte Ca²⁺ uncaging evokes local hyperemia sensitive to COX-1 inhibition, and that neuronal stimulation drives both astrocytic Ca²⁺ elevations and arteriole dilation, effects suppressed by blocking mGluRs or COX-1. ⁷¹ Residual vasodilation found in the presence of COX inhibition suggested that additional pathways could be involved. ⁷¹ Indeed, elevations of intracellular Ca²⁺ in astrocytic endfeet open BK_Ca channels in astrocytic endfeet, leading to K⁺ efflux. ⁷² The resulting extracellular K⁺ increase activates inwardly rectifying K⁺ (K_IR) channels in VSMCs or pericytes, causing hyperpolarization, closure of voltage-dependent Ca²⁺ channels, reduced intracellular Ca²⁺, and vasodilation ⁷² (Figure 3). Combined inhibition of BK channels and COX-1 fully abolished activity-evoked vasodilation, demonstrating that astrocytes contribute to NVC through both COX-dependent prostaglandin and a BK/K⁺-dependent electrical pathway. ⁷² However, the physiological relevance of the latter mechanism remains debated as NVC is normal in mice lacking a BK channel subunit, ⁷³ and several studies question the role of astrocytic mGluRs in this process.^74,75

Additional work indicates that astrocytic Ca²⁺ may not be strictly required for NVC. Selective chemogenetic activation of astrocytes, and sensory-evoked dilation persists in mice lacking inositol 1,4,5-trisphosphate receptor type 2 (IP₃R2), a Ca²⁺-channel activated by inositol trisphosphate (IP₃) mediating the release of Ca²⁺ from the endoplasmic reticulum.^76–79 While IP₃R2 loss abolishes nearly all somatic Ca²⁺ transients, it has a much smaller effect on astrocytic processes, where roughly 40% of Ca²⁺ transients remain. ⁸⁰ Activation of olfactory sensory neuron terminals reliably elicits Ca²⁺ increases in astrocytic processes but not in the soma,^81–83 indicating that localized Ca²⁺ increases in astrocytic endfeet, rather than somatic Ca²⁺ signals, are likely the critical for NVC.

The temporal profile of astrocytic responses further supports a role in amplification rather than in the initiation of the vasodilation induced by neural activity. Ca²⁺ increases in astrocytic cell bodies and processes occur seconds after neuronal activation and the onset of vasodilation.^76,78,84 DREADD-mediated stimulation of astrocytic Gq-GPCR signaling elevates endfeet Ca²⁺ without altering nearby arterioles diameter, ⁷⁸ and whisker stimulation triggers rapid arteriole dilation while endfeet Ca²⁺ rises more slowly. ⁸⁴ Recent evidence reconciles these findings, showing that astrocytes amplify CBF responses when neuronal activity is sustained, even if they do not initiate or mediate the initial rapid dilation. ⁸⁵

Astrocytic contributions also appear vessel-type specific. ³⁷ At capillaries, vasodilation depends on astrocytic Ca²⁺ signaling mediated by P₂X1 purinergic receptors, ³⁷ initiating AA release via phospholipase D and COX-1-dependent prostaglandin E₂ (PGE₂) production, which acts on pericyte EP4 receptors to drive dilation ³⁷ (Figure 3). In contrast to capillary dilation, arteriole dilation does not depend on P₂X1 receptors, PLA₂, PLD₂, or astrocyte Ca²⁺ signaling, but is mainly dependent on neuronal NMDA receptor activation and NO synthesis, as NOS inhibition or NMDA receptor antagonism suppressed the response of arterioles but not capillaries. ³⁷

Astrocytes further contribute to NVC via production of other AA metabolites including epoxyeicosatrienoic acids (EETs). EETs are synthesized from AA by specific cytochrome P450 enzymes epoxygenases, and mediate vasodilation by activating BK_Ca and blocking thromboxane receptors in VSMCs ⁵⁸ (Figure 3). Taken together, these findings indicate that astrocytes play a crucial role in the amplification of the neurovascular response, complementing the rapid initiation by neurons and the retrograde electrical propagation mediated by cECs.

Pericytes: Dynamic and multifaceted regulators of capillary blood flow

The precise role of pericytes in regulating CBF has been a subject of intense debate, a controversy rooted less in irreconcilable data and more in inconsistent definitions of both “pericytes” and “capillaries,” as well as uncertainty about where the arteriolar tree ends and the capillary network begins. ¹³ Because mural cell phenotype changes gradually along the arteriole-capillary continuum, and no single molecular marker cleanly distinguishes pericytes from VSMCs, different studies have often classified the same vessel segments and the same cells in different ways. As a result, much of the apparent conflict in the literature likely reflects differences in anatomical nomenclature rather than true biological disagreement. ¹³ Building on this framework, this section reviews recent in vivo evidence showing that pericytes are not a uniform cell type. Instead, distinct subpopulations are specialized for different tasks, including rapid vasomotion, slow flow modulation, and sensory signal transduction within the neurovascular unit.

The proximal-to-distal transition in pericyte morphology, from the “ensheathing” cells of first-order capillaries (first to third/fourth-order), which feature short processes that nearly completely encase the vessel, to the “mesh” and ultimately the “thin-strand” phenotypes of deeper capillaries (fourth/fifth to ninth-order), characterized by longer, slender processes and increasingly sparse vessel coverage, is now understood to correlate with profound differences in physiological function^6,13 (Figure 1). A central and enduring question arising from this heterogeneity is whether these morphologically distinct pericyte populations are indeed contractile and capable of dynamically modulating capillary diameter to regulate local blood flow.

If capillaries are defined as all vessels between the feeding arteriole and the draining venule, accumulating evidence supports the idea that at least some pericytes are contractile and can modulate capillary diameter both in vitro and in vivo.^{37,51,86–93} Pericytes have been shown to constrict in response to ATP or UTP, ⁸⁸ thromboxane A2 analogs, ⁸⁹ platelet-derived growth factor-BB (PDGF-BB), ⁸⁶ and angiotensin II. ⁸⁷ Conversely, they contribute to the vasodilation evoked by neuronal activity, with astrocyte-derived PGE₂ acting on pericyte EP4 receptors identified as a key mechanism for pericyte relaxation and capillary dilation^37,90 (Figure 3). Consistent with a functional role in NVC, mice with ~30% pericyte loss due to haploinsufficiency of platelet-derived growth factor receptor-β (PDGFR-β), exhibit impaired NVC and reduced brain oxygenation. ⁹⁴

However, this view has been challenged by studies failing to detect a significant role for pericytes in regulating CBF.^89,95,96 For instance, single-cell in vivo optogenetic activation of VSMCs constricts arterioles and reduces flow, while identical activation of pericytes produces no detectable changes in capillary diameter. ⁹⁵ Similarly, sensory stimulation evokes rapid and robust dilation of VSMC-covered vessels but only slow and nearly undetectable dilation of pericyte-covered capillaries. ⁹⁵ As noted earlier, many of these apparent contradictions likely stem from inconsistent definitions of where the arteriole ends and the capillary begins, leading different studies to classify the same transitional vessel segments, and their associated mural cells, in different ways. ¹³

To conclude that distal pericytes are entirely non-contractile, however, would be an oversimplification. Instead, a more nuanced picture is emerging, one in which pericyte contractility is not absent but kinetically distinct along the vascular tree. Hartmann et al., ⁹³ employed two-photon laser ablation to selectively remove individual distal capillary pericytes (fifth to ninth order), without disrupting the blood–brain barrier, observing an increase in blood flow in the affected capillary segments. ⁹³ Crucially, the same study found that the relaxation kinetics of these distal pericytes in response to neuronal stimulation are an order of magnitude slower than the rapid dilation of ensheathing pericytes (first to third/fourth-order). ⁹³ This positions ensheathing pericytes as the early responders contributing to the initial, rapid phase of NVC, while the slow relaxation kinetics of distal capillary pericytes are more consistent with a role in the late phases of NVC. ⁹³ The novelty here lies in shifting the paradigm from a binary “contractile versus non-contractile” debate to a more nuanced model of “fast versus slow” regulation. This conclusion was reinforced by the evidence that, depending on their location, pericytes respond differently to vasoconstrictive agents: proximal pericytes contract in response to the thromboxane A2 agonist U46619 or to a depolarizing concentration of K⁺ (60 mM). In contrast, distal pericytes also contract in response to U46619, but they do not respond to 60 mM K⁺, and their contraction occurs much more slowly than that of proximal pericytes. ⁹² Taken together, these findings further suggest that capillary pericytes can be broadly categorized into arteriole-proximal and distal populations, which regulate blood flow on fast and slow time scales, respectively, reflecting the corresponding kinetics of their contractile responses. ¹³

Further refining pericytes role in NVC, recent work has demonstrated that distal pericytes are not merely slow effectors but can also serve as sensory and signal-propagating nodes. They appear to sense neural activity and convert it into K_ATP -dependent electrical signals that propagate retrogradely to upstream arterioles, enabling precise spatiotemporal regulation of blood flow ⁹⁷ (Figure 4). This is consistent with the findings of Gonzales et al., ⁹² who demonstrated that pericytes in the post-arteriole transitional region relax in vivo in response to K_IR-dependent retrograde hyperpolarizing signals triggered by local K⁺ application to downstream capillaries, highlighting the direct electrical communication between capillary endothelial cells and pericytes via intercellular junctions. ⁹² Collectively, these findings suggest that thin-strand pericytes, like capillary ECs,^42,62 may function primarily as sensors and signal transducers within the neurovascular unit, rather than as the principal effectors of rapid diameter changes.

Microglia and purinergic modulation

Although often overlooked as a component of the NVU, recent evidence has demonstrated that microglia may also actively regulate CBF.^9–11 It has been estimated that about 30% of microglia observed in adult mice are associated with cerebral blood vessels ⁹ and establish direct contacts with VSMCs, pericytes, and astrocytes, cell types involved in CBF regulation. ¹⁰ Microglia can also interact directly with endothelial cells through gaps between adjacent astrocytic endfeet.^10,98–100 Supporting their role in CBF regulation, depletion of microglia has been associated with a reduced CBF response to whisker stimulation and an increase in capillary diameter.^9,10 Although the precise mechanisms remain unclear, evidence showing that functional hyperemia is suppressed following microglial P₂Y₁₂ receptor (P₂Y₁₂R) deletion suggests that microglial purinergic signaling may play a role in mediating their ability to regulate CBF.^10,11 Recent work has provided additional mechanistic insight by showing that the ectonucleotidase CD39, which initiates the dephosphorylation of extracellular ATP into adenosine, also contributes to microglia-dependent CBF regulation ¹¹ (Figures 2 and 3). Pharmacological inhibition or microglia-specific deletion of CD39 reproduces the CBF alterations observed in microglia-deficient mice. ¹¹ Overall, these findings indicate that microglia may use P₂Y₁₂R to sense extracellular ATP and orient their processes to initiate CD39-mediated breakdown of ATP into adenosine, which emerges as a crucial mediator of microglia-dependent regulation of cerebrovascular reactivity.

Angiotensin II, perivascular macrophages, and oxidative stress

A substantial body of evidence implicates angiotensin II (Ang II) signaling as a central driver of neurovascular dysfunction in hypertension. Acute or chronic administration of angiotensin-II (Ang II) attenuates NVC, ¹¹⁴ an effect, that is, rescued by reactive oxygen species (ROS) scavenging, indicating a key role for vascular oxidative stress. ¹¹⁵ Several lines of evidence suggest that this impairment is independent of systemic BP elevation. First, in the slow-pressor Ang II model of HTN, NVC, and endothelium-dependent vasodilation deficits precede the onset of elevated BP and persist even after normalization of BP. ¹¹⁶ Second, topical neocortical application of Ang II reproduces these deficits without altering systemic BP. ¹¹⁴ Third, CBF responses remain intact when BP is increased by phenylephrine, a vasoconstrictor that does not activate Ang II signaling. ¹¹⁴ Consistent with these observations, genetic models of HTN, characterized by elevated circulating Ang II, including BPH mice ¹¹⁷ and spontaneously hypertensive rats (SHR), ¹¹⁸ also exhibit impairment of CBF responses. Mechanistically, Ang II does not appear to act directly on neurons, as it leaves whisker-evoked field potentials unchanged after either acute ¹¹⁴ or chronic ¹¹⁶ administration. Instead, Ang II activates AT1R in perivascular macrophages (PVMs), innate immune cells closely associated with cerebral arterioles and venules ¹¹⁷ (Figure 5). Activation of AT1R signaling in PVMs leads to increased reactive oxygen species (ROS) production via nicotinamide adenine dinucleotide phosphate (NADPH) oxidase 2 (NOX2), most likely resulting in NO scavenging and consequent impairment of endothelium-dependent vasodilation and NVC ¹¹⁷ (Figure 5). Consistent with this, depletion of PVMs by intracerebroventricular clodronate or AT1R deletion in bone-marrow derived cells prevents the neurovascular dysfunction. ¹¹⁷ Additional evidence from stroke-prone spontaneously hypertensive rats (SHRSP) further supports this concept. ¹¹⁹ The relative contribution of other cell types remains incompletely defined. Although brain ECs also express AT1R and NOX2, ¹²⁰ single-cell RNA sequencing indicates that NOX2 expression in ECs is relatively low compared with microglia and PVMs, ¹²⁰ suggesting that ECs are unlikely to be the primary source of Ang II-induced oxidative stress. Nevertheless, recent evidence points to additional, AT1R-independent mechanisms of endothelial dysfunction in HTN, including impaired function of endothelial K_IR2.1 channels, ¹²¹ indicating that multiple parallel pathways may contribute.

OPEN IN VIEWER

Figure 5.

Mechanisms of cerebrovascular dysfunction in hypertension. In ANGII slow-pressor hypertension, increased BBB permeability permits circulating ANGII to enter the perivascular space and activate AT1R on BAMs. This, in turn, stimulates NOX2-dependent ROS production, resulting in cerebrovascular dysfunction via scavenging of endothelial- and neuronal-derived NO and consequent formation of ONOO⁻. In the DOCA-salt model, cerebrovascular dysfunction depends on two distinct IL-17-dependent mechanisms involving brain ECs and BAMs: (1) in the circulation, IL-17 produced by intestinal T cells acts on endothelial IL-17RA, reducing endothelial NO production and suppressing endothelium-dependent vasodilation and (2) in the brain, IL-17 produced by dura-associated γδ-T cells acts on IL-17RA expressed by BAMs, inducing vascular oxidative stress and impairing functional hyperemia, with minimal effects on endothelial vasodilatory function. Dashed line indicates blunted eNOS-derived NO production due to IL-17A-mediated inhibition.

IL-17: A unifying link between hypertension, high salt, and neurovascular dysfunction

A distinct pathway involving interleukin-17 (IL-17) has emerged from studies of both dietary salt loading and the deoxycorticosterone acetate (DOCA)-salt model of hypertension, revealing a common immune-mediated mechanism that operates independently of classical Ang II signaling.

Excessive dietary salt intake has emerged as an independent risk factor for cerebrovascular disease and cognitive impairment, distinct from its effects on BP (reviewed by Faraco ¹²² ). In mice, high salt diet (HSD) reduces resting cerebral perfusion and selectively impairs endothelial NO-dependent vasodilation while leaving NVC intact. ¹²³ These effects are mediated by a gut-derived Th17 immune response that increases circulating IL-17A, which in turn suppresses eNOS activity via inhibitory phosphorylation at Thr⁴⁹⁵ through Rho-kinase activation in brain ECs. ¹²³ Genetic deletion or pharmacological neutralization of IL-17A prevents the cerebrovascular effects of HSD, while systemic IL-17A administration recapitulates them. ¹²³

Building on this foundation, subsequent work in the DOCA-salt model of HTN revealed that IL-17 also contributes to neurovascular dysfunction, but with distinct mechanistic features. ¹²⁴ In this model, neurovascular dysfunction is not mediated by central Ang II signaling but rather by IL-17 acting on both sides of the blood–brain barrier (BBB). ¹²⁴ Importantly, IL-17 acts on distinct cellular targets on either side of the BBB. Circulating IL-1, likely released by intestinal T cells, acts on endothelial IL-17 receptor A (IL-17RA) to reduce NO production via inhibitory phosphorylation of eNOS at Thr^495,123,124 impairing endothelium-dependent vasodilation without affecting NVC ¹²⁴ (Figure 5). In contrast, IL-17 derived from dura-associated T cells acts on IL-17RA expressed by border-associated macrophages (BAMs), inducing vascular oxidative stress and suppressing NVC, with minimal effects on endothelial function ¹²⁴ (Figure 5). These findings reveal that the same cytokine can produce distinct functional deficits depending on which side of the BBB it engages, highlighting the complexity of neurovascular regulation in disease.

A critical question is whether these IL-17-mediated cerebrovascular deficits have functional consequences for the brain. Both the HSD and DOCA-salt models are associated with cognitive impairment,^123,124 and a mechanistic link between endothelial dysfunction and cognitive decline has been proposed. ¹²⁵ Faraco et al. ¹²⁵ demonstrated that HSD promotes tau phosphorylation and aggregation through a pathway originating from reduced eNOS-derived NO. ¹²⁵ Specifically, endothelial NO deficiency suppresses calpain nitrosylation, leading to calpain activation, cleavage of p35–p25, overactivation of cyclin-dependent kinase 5 (CDK5), and subsequent pathological tau phosphorylation. ¹²⁵ The functional relevance of this pathway is supported by evidence that tau-null mice are protected from HSD-induced cognitive deficits, as are animals treated with anti-tau antibodies. ¹²⁵ While this mechanism remains to be validated in the DOCA-salt model, it illustrates how disruption of CBF regulation, specifically endothelial NO deficiency driven by IL-17, can directly impact neuronal function and contribute to cognitive impairment. Whether these mechanisms translate to humans remains largely unknown. However, circulating levels of IL-17A have been shown to correlate with cognitive impairment,^126–128 suggesting that these pathways may be clinically relevant.

Central pathways and downstream effectors

Beyond local vascular mechanisms, central pathways may also contribute to hypertension-induced dysfunction. In the slow-pressor Ang II model, cerebrovascular impairment requires ROS production in the subfornical organ (SFO) and involves not only AT1R activation but also vasopressin release, upregulation of endothelin-1 (ET-1) in cerebral arterioles, and activation of ET-1 type A (ET_A) receptors. ¹¹¹ Both ET-1 and Ang II appear to converge on vascular oxidative stress as a common downstream mechanism. ¹¹¹ Because ET_A receptors are highly expressed in pericytes, ¹²⁰ it is plausible that pericyte-mediated vascular dysfunction may contribute to the observed impairment of NVC, although direct evidence is lacking. More broadly, recent transcriptomic analyses have revealed major molecular changes across multiple NVU cell types in HTN, providing the foundation for further mechanistic studies aimed at defining how coordinated gene expression alterations drive neurovascular dysfunction. ¹²⁹

Amyloid-β (Aβ)

The vasoactive properties of Aβ were first demonstrated by Thomas et al., ¹³⁴ who showed that Aβ constricts isolated aortas and impairs endothelium-dependent relaxation to acetylcholine, an effect attributed to increased vascular oxidative stress. ¹³⁴ Subsequent work has confirmed that endothelial dysfunction is present in cerebral blood vessels of APP transgenic mice, ¹³⁵ while response to NO donors remained intact, indicating that the impairment is not due to a nonspecific loss of vascular smooth muscle responsiveness. ¹³⁵ Later studies in Tg2576 mice revealed that Aβ disrupts multiple regulatory processes governing CBF, including NVC, endothelial function, and CA.^136–139 Importantly, superfusion of the somatosensory cortex of wild-type mice with Aβ_1–40, but not Aβ_1–42, recapitulates the impairments observed in APP transgenic mice, suggesting that different Aβ peptides exert distinct vasoactive effects.^140,141 Cerebrovascular dysfunction has been reported across multiple APP transgenic models and brain regions,^142–151 with vascular abnormalities typically emerging before amyloid plaque deposition, tau pathology, and cognitive decline, implicating neurovascular dysfunction as an early pathogenic event in AD. ¹⁵²

A unifying theme across these studies is the central role of oxidative stress. Endothelial dysfunction in APP-overexpressing mice can be rescued by exogenous superoxide dismutase (SOD) or prevented by SOD1 overexpression, ¹³⁶ and the free radical scavenger MnTBAP counteracts the effects of Aβ_1–40 on resting CBF. ¹⁴¹ Mechanistic studies have identified CD36 as a critical molecular link between Aβ_1–40 and vascular oxidative stress. The cerebrovascular effects of Aβ are absent in Tg2576 mice lacking CD36 and in CD36 knockout mice exposed to exogenous Aβ_1–40. ¹⁵³ Subsequent work has demonstrated that Aβ activates CD36 specifically on BAMs, triggering NOX2-dependent superoxide production, which scavenges NO and impairs both NVC and endothelium-dependent vasodilation ¹⁵⁴ (Figure 6). Consistent with this, depletion of BAMs or transplantation of CD36^−/− bone-marrow cell rescues neurovascular function. ¹⁵⁴ Because PVMs are predominantly located along penetrating arteries and precapillary arterioles, these findings suggest that Aβ primarily disrupts neurovascular regulatory mechanisms at this level of the cerebrovascular tree.

OPEN IN VIEWER

Figure 6.

Examples of mechanisms underlying cerebrovascular dysfunction in AD at multiple levels of the cerebral microvasculature. At the arteriolar level, activation of CD36 on PVMs triggers NOX2-dependent production of ROS, leading to vascular oxidative stress. This, in turn, scavenges NO, resulting in impaired endothelium-dependent vasodilation and NVC. At the capillary level, activation of NOX4 in pericytes induces oxidative stress, which increases ET-1 expression in brain ECs. ET-1 then acts on ET_AR on pericytes, causing their contraction and reducing capillary diameter and CBF. The resulting reduction in CBF compromises the delivery of oxygen and glucose, which can activate BACE1, thereby increasing Aβ production. This establishes a vicious cycle in which vascular dysfunction promotes amyloidogenesis, further exacerbating cerebrovascular and neuronal dysfunction. In parallel, tau pathology impairs NVC by disrupting the functional coupling between NMDAR, PSD95, and nNOS, thereby reducing nNOS-derived NO production and further limiting NO bioavailability, which contributes to impaired NVC.

However, more recent evidence indicates that capillaries are also highly vulnerable to the cerebrovascular effects of Aβ. In both humans with developing AD and in App^NL-G-F mice, capillaries are selectively constricted, whereas arteriolar and venular diameters remain unchanged. ¹¹² Mechanistically, Aβ directly constricts capillaries through a pathway involving NOX4-dependent ROS production, which triggers endothelin-1 (ET-1) release and subsequent activation of endothelin type A (ET_A) receptors on pericytes, inducing contraction and capillary narrowing. ¹¹² This capillary-specific pathway appears distinct from the PVM-dependent mechanism at the arteriolar level, suggesting that Aβ disrupts different segments of the vascular tree through cell-type-specific mechanisms. Importantly, Aβ-evoked capillary constriction can be prevented by NOX4 or ETA receptor inhibition and reversed by the vasodilator C-type natriuretic peptide ¹¹² or the voltage-gated calcium channel blocker nimodipine. ¹⁵⁵ The resulting reduction in CBF and energy supply may, in turn, increase Aβ production by upregulating β-amyloid precursor protein-cleaving enzyme 1 (BACE1), establishing a vicious cycle ¹¹² (Figure 6). A broader role for ET-1 in AD-related cerebrovascular dysfunction is further supported by evidence that loss of CD2-associated protein (CD2AP), an endothelial protein whose levels are reduced in AD brains, ¹⁵⁶ impairs NVC, resting CBF, and vasomotion, effects partially rescued by antagonism of ET-1 signaling. ¹⁵⁶

The importance of oxidative stress in mediating the vascular effects of Aβ is further underscored by another study showing that Aβ-induced oxidative stress promotes DNA damage, activation of poly(ADP-ribose) polymerase (PARP), and ADP-ribose-dependent opening of transient receptor potential melastatin 2 (TRPM2) channels, leading to elevated intracellular Ca²⁺ levels and endothelial dysfunction. ¹⁵⁷ Together with the previous evidence, these findings support a model in which oxidative stress is initially triggered in PVMs via CD36–NOX2 signaling and subsequently propagates to brain ECs, where it promotes a cascade of events that amplifies vascular dysfunction.

Beyond oxidative stress, Aβ can compromise endothelial function through other pathways. Specifically, studies in 5xFAD and APP23 mouse models have shown impairments in both functional hyperemia and K⁺-induced retrograde vasodilation,^146,158 effects driven by a suppression of K_IR2.1 channel activity in cECs.^146,158 Notably, systemic administration of PIP₂, a critical regulator of K_IR2.1 channels, successfully restores channel function and rescues the cerebrovascular alterations. ¹⁴⁶ A separate pathway involving reduced tissue plasminogen activator (tPA) activity, resulting from increased plasminogen activator inhibitor-1 (PAI-1) activity, has been implicated in NVC impairment in AD. ¹⁵⁹ Because tPA modulates NMDAR activity and couples neuronal activation to nNOS-derived NO production, reduced tPA activity may uncouple this signaling cascade, selectively impairing NVC without affecting endothelial function. ¹⁵⁹ This interpretation is supported by evidence that Aβ does not impair NVC in tPA- or PAI-1-null mice. ¹⁵⁹ Additionally, mTOR-mediated inhibition of nNOS phosphorylation has been proposed as another mechanism reducing nNOS-derived NO and attenuating NVC in AD. ¹⁴⁷ Interestingly, although most studies do not report alterations in VSMC function,^{136,144,149,153,154,157,159,160} a recent study found that BK_Ca channel activity, which normally opposes vasoconstriction via hyperpolarizing currents in VSMCs, is reduced in 5xFAD mice via sex-dependent mechanisms, ¹⁶¹ suggesting that this reduction may contribute to a hypercontractile state in the cerebral microcirculation in AD. Together, these observations indicate that Aβ disrupts cerebrovascular regulation through multiple parallel pathways that target different cell types and vessel segments. An important open question is the relative contribution of each pathway at different stages and whether they represent independent or interacting processes.

Tau

Tau pathology has also been implicated in neurovascular uncoupling, ¹⁶² though through mechanisms distinct from those engaged by Aβ. In PS19 mice overexpressing human tau, NMDAR-dependent NO release and NVC are markedly reduced, and these deficits precede overt neurodegeneration or tangle formation. ¹⁶² In the rTg4510 mice, doxycycline-induced suppression of tau production rescues NVC, ¹⁶² whereas neurofibrillary tangles formation is unaffected, pointing to soluble tau species, rather than insoluble tangles, as the mediators of vascular dysfunction. ¹⁶² Mechanistically, tau disrupts the interaction between nNOS and its scaffolding protein PSD95, uncoupling NMDAR activation from neuronal NO production ¹⁶² (Figure 6). Whether this mechanism interacts with the mTOR-mediated suppression of nNOS expression also reported in AD ¹⁴⁷ remains to be determined.

The effects of tau on endothelial function are more complex and somewhat inconsistent across studies. Park et al. ¹⁶² found that endothelium-dependent vasodilation to acetylcholine, bradykinin, or the Ca²⁺ ionophore A23187 was intact in both PS19 or rTg4510 mice, ¹⁶² and topical application of exogenous tau to wild-type cortex did not impair endothelial responses. ¹⁶² However, resting CBF, which is largely dependent on endothelial NO, was reduced in PS19 mice. ¹⁶² More recent studies using genetically encoded Ca²⁺ biosensors have shown that soluble pathogenic tau peptides disrupt endothelial Ca²⁺ signaling required for NO production and vasodilation, ¹⁶³ and accumulation of tau in brain ECs disrupts eNOS trafficking, leading to eNOS inhibition and impaired NO-mediated vasodilation. ¹⁶⁴ These discrepancies may reflect differences between acute and chronic tau exposure, distinct tau species, or the specific vascular beds examined. Resolving these inconsistencies will require systematic studies directly comparing endothelial function across models and conditions.

ApoE4 genotype

The ApoE4 genotype, the strongest genetic risk factor for sporadic AD, ¹⁶⁵ increases susceptibility to microvascular pathology, including white matter damage and amyloid-related imaging abnormalities (ARIA) observed during anti-Aβ immunotherapy. Humanized ApoE4 knock-in mice (ApoE4-TR) exhibit marked impairment in NVC,^166,167 dependent on ROS production via NADPH oxidase.^166,167 As in models of hypertension and Aβ accumulation,^117,154 BAMs are the principal pathogenic source of ROS. ¹⁶⁷ Depletion of BAMs or selective deletion of ApoE4 in BAMs using Mrc1^Cre+/ApoE4^fl/fl mice fully rescues neurovascular function without altering total brain or circulating ApoE levels, identifying BAM-derived ApoE4 as the key driver of cerebrovascular dysfunction. ¹⁶⁷ This striking convergence with the Aβ literature, both involving perivascular macrophages and oxidative stress, suggests that BAMs may represent a common cellular target of diverse AD-related risk factors and a critical therapeutic target. However, whether interventions targeting BAMs would be effective once dysfunction is established, and whether they would translate to humans, remains to be determined. Additional evidence indicates that ApoE4 expression in vascular mural cells also contributes to dysfunction. Expression of ApoE4, but not ApoE3, in vascular mural cells impairs arteriolar blood flow, ¹⁶⁸ pointing to pericytes and VSMCs as other critical cellular mediators of ApoE4 vascular toxicity. Whether this reflects direct effects on mural cell function or interactions with adjacent endothelial cells remains to be determined.